TWI392986B - System and method for integrating dispersed point-clouds of an object - Google Patents

Description

Discrete point cloud integration system and method

The present invention relates to a point cloud processing system and method, and more particularly to a system and method for integrating discrete points obtained by scanning the same object multiple times.

Reverse engineering is relative to forward engineering. The so-called forward engineering refers to the design drawings of existing products, and then the physical products are processed according to the drawings. Reverse engineering is a high-speed three-dimensional laser scanner that accurately and high-speed scans existing physical objects (samples or models), acquires point cloud data of real objects, constructs a three-dimensional digital model based on the obtained point cloud data, and then uses the CAM system to complete Manufacturing of products.

Scanning the same object with a laser scanner is generally not possible to scan at one time. To obtain a complete set of point cloud data of an object, it is usually necessary to scan the object from different angles (multi-view data acquisition), and then use some methods to The discrete point clouds acquired by the sub-scan are restored to their original spatial positional relationship by alignment reset, and merged into a complete point cloud to obtain a complete three-dimensional information of the surface of the object.

At present, such methods in the industry have special requirements for the collection process, or strict control of sophisticated and complex displacement devices. For example, the method of positioning by the characteristics of the product (dotted line on the product): presetting the marked points on the product, according to the relationship between the number of marked points in the three-dimensional data of each two different perspectives obtained by scanning. , to determine the coordinate conversion relationship between the various perspectives. However, multiple positioning will produce a large error. The high precision operation is very troublesome.

In view of the above, it is necessary to provide a discrete point cloud integration system and method, and the discrete point clouds acquired by multiple scanning objects are restored to their original spatial positional relationship by alignment reset, and merged into one complete point cloud. To obtain a more complete three-dimensional information on the surface of the object.

A discrete point cloud integration system includes a computer, a scanner coupled to the computer, and a fixture for placing the scanned object, the fixture having three small balls thereon. The computer includes: a point cloud acquisition module, configured to acquire a point cloud of an object obtained by the scanner scanning each side of the object, and a point cloud corresponding to three small balls on the jig of each side; a ball fitting module, A set of small balls is respectively fitted to the point clouds of the three small balls obtained from each side of the scanned object; a calculation module is used to calculate the position of each of the small balls in each group, and calculate each group Fit the distance between each two small balls in the ball; the matching module is used to use one side of the object as a reference surface, and the matching small ball obtained by scanning the surface is used as a reference ball to find other faces and The small ball matching the small balls on the surface; and the alignment module is used to align the matched balls of the other groups obtained by scanning the other faces according to the matching relationship, and the aligned ball is aligned as a reference. Position, obtain the transformation matrix in the alignment process, and use the point cloud of the object obtained by scanning the reference plane of the object as a reference point cloud, and align the point cloud obtained by scanning each surface of the object to the reference point cloud according to the obtained transformation matrix. Position, get the object finished Point cloud.

A discrete point cloud integration method, the method comprising the steps of: (A) fixing an object to be scanned on a jig that can be inverted and containing three small balls; (B) flipping the jig to scan all faces of the object, in scanning Simultaneously, the three small balls on the jig are scanned once on each side of the object; (C) acquiring a point cloud of the object obtained by scanning each side of the object, and three small balls on the jig obtained by scanning the object on the side a point cloud; (D) fitting a set of small balls according to the point clouds of the three small balls obtained on each side of the scanned object; (E) calculating the position of each small ball in each set of fitted small balls, and calculating Each group fits the distance between each two small balls in the small ball; (F) uses one side of the object as a reference surface to scan the surface of the object The obtained fitting ball is a reference ball, and the small ball matching the small balls of the surface in other faces is found; (G) each group of fitting balls obtained by scanning other faces of the object is translated according to the matching relationship, The rotation transformation is aligned to the position of the fitted sphere as the reference to obtain the transformation matrix in the alignment process; and (H) the point cloud of the object obtained by scanning the object reference plane is used as the reference point cloud, according to the obtained transformation matrix The point cloud obtained by scanning each side of the object is aligned to the reference point cloud to obtain a complete point cloud of the object.

Compared with the prior art, the discrete point cloud integration system and method provided by the present invention utilizes three small balls to align and reset the discrete point clouds acquired by multiple scanning objects, and restore the original spatial positional relationship between the two. The complete point cloud of the object, simple operation and high precision.

10‧‧‧ fixture

20‧‧‧Laser scanner

30‧‧‧ computer

310‧‧‧Point Cloud Acquisition Module

320‧‧‧Ball fitting module

330‧‧‧ Point Cloud Trim Module

340‧‧‧Computation Module

350‧‧‧ Matching module

360‧‧‧Alignment module

370‧‧‧ point cloud output module

1 is a hardware architecture diagram of a preferred embodiment of a discrete point cloud integration system of the present invention.

Figure 2 is a functional block diagram of the computer of Figure 1.

3 is a flow chart of a preferred embodiment of the discrete point cloud integration method of the present invention.

Fig. 4(A), Fig. 4(B), and Fig. 4(C) are schematic diagrams showing different stages of the two sets of fitting ball alignment processes.

As shown in FIG. 1, it is a hardware architecture diagram of a preferred embodiment of the discrete point cloud integration system of the present invention. The system mainly includes a jig 10, a scanner (in this embodiment, a laser scanner 20), and a computer 30.

The jig 10 is used to place an object A to be scanned, and the jig 10 has three ceramic balls, as shown in the figure, a ball a, a ball b and a ball c. The size of the three small balls may be the same or different. In this embodiment, the three small balls are the same size, and the triangle formed by the three small balls is an equilateral triangle. The jig can be flipped 360° to facilitate scanning of the various faces of object A.

The laser scanner 20 is configured to scan the object A placed on the jig 10 to obtain a discrete point cloud of the object A obtained by multiple scans. While scanning each side of the object A, the laser scanner 20 will correspond to the face. The three small balls on the ball are scanned once to obtain a point cloud of three small balls.

The computer 30 is configured to receive a point cloud of the object A obtained by scanning the laser A on each side of the object A and a point cloud of three small balls, and respectively fit the point clouds of the small balls obtained by scanning each side of the object A (fit) ) a small ball, such as: scan the front of the object A to get the point cloud "scan0", "scan1", "scan2" of the small ball a, b, c, using the least squares method according to the point cloud "scan0", "scan1", " Scan2" fits a set of small balls Q1, Q2, Q3; flips the fixture 10, scans the opposite side of the object A to get the point cloud "scan3", "scan4", "scan5" of the small ball a, b, c, using the smallest two Multiplication is based on the point cloud "scan3", "scan4", "scan5" to fit a set of small balls M1, M2, M3. After that, the computer 30 matches the fitted small balls, for example, the fitted small ball M1 corresponds to Q1, M2 corresponds to Q2, and M3 corresponds to Q3.

Next, the computer 30 selects a certain surface of the object A as a reference surface, and scans the matching sphere obtained by the surface as a reference ball, and aligns the other groups of the small spheres with the transformation and rotation to align with the reference sphere. (align) to each matching small ball coincides to obtain a transformation matrix. Further, the computer 30 uses the point cloud of the object A obtained by scanning the reference plane of the object A as a reference point cloud, and aligns the point cloud obtained by the other surface of the scanned object A with the reference point cloud according to the obtained transformation matrix, thereby obtaining the object A intact. Point cloud.

As shown in FIG. 2, it is a functional module diagram of the computer 30 in FIG. The computer 30 includes a point cloud acquisition module 310, a ball fitting module 320, a point cloud clipping module 330, a calculation module 340, a matching module 350, an alignment module 360, and a point cloud output module 370.

The point cloud acquisition module 310 is configured to acquire a point cloud of the object A obtained by the laser scanner 20 scanning each side of the object A and a point cloud corresponding to the three small balls on the jig of each side.

The ball fitting module 320 is configured to respectively fit a set of small balls according to the point clouds of the three small balls obtained by scanning each side of the object A, for example, the point cloud of the small objects a, b, c is obtained by scanning the front of the object A. Scan0"," Scan1", "scan2", using the least squares method to fit a set of small balls Q1, Q2, Q3 according to the point cloud "scan0", "scan1", "scan2"; flip the fixture 10, scan the object A to get the small ball a, b, c point cloud "scan3", "scan4", "scan5", using the least squares method to fit a set of small balls M1, M2, M3 according to the point cloud "scan3", "scan4", "scan5" .

The point cloud trimming module 330 is further configured to trim the chaotic point cloud of each fitting ball, and is used to delete the fitting ball and the jig obtained by scanning each surface of the object after obtaining the complete point cloud of the object A. Three small balls of point clouds.

The calculation module 340 is configured to calculate the position of each of the small balls in each set of fitted balls, and calculate the distance between each of the two small balls in each set of fitted balls.

The matching module 350 is configured to use a certain surface of the object A as a reference surface, and the matching small ball obtained by scanning the surface is used as a reference small ball to find a small ball matching the small balls of the surface in the other surface. Suppose that a set of fitting balls Q1, Q2, and Q3 obtained from the front side of the scanning object A are used as reference balls, because the distance between the two small balls a, b, and c on the jig 10 is not equal: | ab | ≠| bc |≠| ca |, so the distance between two or two small balls in each set of fitting balls is not equal, and | Q 1 Q 2|≠| Q 2 Q 3|≠| Q 3 Q 1| , M 1 M 2|≠| M 2 M 3|≠| M 3 M 1|, according to the distance between two small balls in each set of fitting balls, and the other groups fit two or two small balls in the small ball Correspondence between the distances, find the correspondence between the small balls of each group of fitting balls, if | Q 1 Q 2|=| M 1 M 2|, | Q 2 Q 3|=| M 2 M 3|,| Q 3 Q 1|=| M 3 M 1|, then it can be concluded that M1 corresponds to Q1, M2 corresponds to Q2, and M3 corresponds to Q3.

The alignment module 360 is configured to match each group of the obtained small balls obtained by scanning other faces of the object A according to the matching relationship. Align to the fitted sphere as a reference, such as fitting small balls M1, M2, M3 corresponding to the fitting balls Q1, Q2, Q3, respectively, to fit the small balls Q1, Q2, Q3 as a reference, the fitting will be small The balls M1, M2, and M3 are transformed as a whole through the translation and rotation until the fitting balls M1, M2, and M3 are aligned with the fitting balls Q1, Q2, and Q3, respectively, to obtain a transformation matrix in the alignment process.

Next, the alignment module 360 uses the point cloud of the object A obtained by scanning the reference plane of the object A as a reference point cloud, and aligns the point cloud obtained by scanning each surface of the object A to the position where the reference point cloud is located according to the obtained transformation matrix. , get the complete point cloud of object A.

The point cloud output module 370 is used to output a complete point cloud of the object A obtained after alignment, and report the alignment accuracy.

As shown in FIG. 3, it is a flowchart of a preferred embodiment of the discrete point cloud integration method of the present invention.

First, the object A is fixed to the jig 10 for auxiliary scanning, and the jig 10 has three ceramic balls a, b, c, and the jig 10 can be turned 360° (step S10).

Next, the jig 10 is turned over, and the laser scanner 20 scans each side of the object A. While scanning each side of the object A, the laser scanner 20 scans the three balls on the jig once (step S12).

The point cloud acquisition module 310 acquires a point cloud of the object A obtained by scanning each surface of the object A, and when scanning each surface of the object A, a point cloud corresponding to three small balls on the jig 10 obtained on the surface, such as Scanning the front side of object A to get three small balls of point cloud as "scan0", "scan1", "scan2", and scanning the object A to get the three small balls of point cloud as "scan3", "scan4", "scan5" ( Step S14).

The ball fitting module 320 applies a mathematical rule, such as a least squares method, to respectively fit a set of small balls corresponding to the face of the object A according to the point clouds of the three small balls obtained on each side of the scanned object A, for example: using the smallest The second multiplication method, fitting a set of small balls according to the point cloud "scan0", "scan1", "scan2" Q1, Q2, and Q3, a set of small balls M1, M2, and M3 (shown in FIG. 4(A)) are fitted according to the point cloud "scan3", "scan4", and "scan5" (step S16).

The point cloud trimming module 330 trims the scrambled point clouds of the respective small balls (step S18).

The calculation module 340 calculates the position of each of the small balls in each set of fitted balls, and calculates the distance between the two small balls in each set of fitted balls, as calculated | Q 1 Q 2|, | Q 2 Q 3|, | Q 3 Q 1|, and calculation | M 1 M 2|, | M 2 M 3|, | M 3 M 1| (step S20).

The matching module 350 takes a certain surface of the object A as a reference surface, and uses the fitted small ball obtained by scanning the surface as a reference ball to find a small ball matching the small balls of the surface in the other surface. Suppose that the front side of the object A is used as a reference surface, and a set of fitting balls Q1, Q2, and Q3 obtained by scanning the front surface of the object A are used as reference balls, because the three small balls a, b, and c on the jig 10 are two or two. The distance between them is not equal: | ab |≠| bc |≠| ca |, so the distance between two or two small balls in each group of fitting balls is not equal, and | Q 1 Q 2|≠| Q 2 Q 3|≠| Q 3 Q 1|,| M 1 M 2|≠| M 2 M 3|≠| M 3 M 1|, according to the distance between the two balls in each set of fitting balls and other groups Fit the corresponding relationship between the distances between the two small balls in the ball, and find the correspondence between the small balls between the fitted balls, assuming that | Q 1 Q 2|=| M 1 M 2|, | Q 2 Q 3|=| M 2 M 3|,| Q 3 Q 1|=| M 3 M 1|, it can be concluded that M1 corresponds to Q1, M2 corresponds to Q2, and M3 corresponds to Q3 (step S22).

The alignment module 360 is configured to align each set of fitting balls obtained by scanning other faces of the object A according to the matching relationship to the fitting ball as a reference, such as fitting the small balls M1, M2, and M3 respectively corresponding to the fitting. Ball Q1, Q2, Q3, can fit the ball Q1, Q2, Q3 as the benchmark, will fit the ball M1, M2, M3 As a whole, the translation and rotation transformations are performed until the fitting balls M1, M2, M3 and the fitting balls Q1, Q2, and Q3 are aligned to coincide, respectively, and the transformation matrix in the alignment process is obtained, and the specific method is as follows: (a) The spatial triangle M1M2M3 composed of the small balls M1, M2, and M3 coincides with the vertex Q1 and the vertex Q1 (as shown in FIG. 4(B)), and the first transformation matrix is obtained; (b) the vertex Q1 is used as the rotation origin and the plane Q1Q2M2 (or plane M1Q2M2, because vertex M1 coincides with vertex Q1), the normal vector is the rotation axis, the inner angle of the edge Q1Q2 and the edge Q1M2 is the rotation angle, and the edge M1M2 is rotated to coincide with the edge Q1Q2 to obtain the second transformation matrix; (c) With the vertex Q1 as the rotation origin, Q1Q2 as the rotation axis, and the inner angle of the side Q1M2 and the side Q1Q3 as the rotation angle, the side M1M3 is rotated to coincide with the side Q1Q3 (as shown in Fig. 4(C)), and the Three transformation matrices (step S24).

The alignment module 360 uses the point cloud of the object A obtained by scanning the reference plane of the object A as a reference point cloud, and aligns the point cloud obtained by scanning the other surface of the object A to the reference point cloud according to the obtained transformation matrix, and obtains the complete object A. Point cloud (S26).

The point cloud trimming module 330 further deletes the point cloud of the three small balls on the jig obtained by fitting the small balls and scanning the respective faces of the object A (S28).

Finally, the point cloud output module 370 outputs the complete point cloud of the object A obtained after the alignment, and reports the alignment accuracy (step S30).

The present invention has been described above in terms of preferred embodiments, and is not intended to limit the invention. The scope of the present invention is defined by the scope of the appended claims, unless otherwise claimed.

10‧‧‧ fixture

20‧‧‧Laser scanner

30‧‧‧ computer

Claims (7)

A discrete point cloud integration system, the system comprising a computer, a scanner connected to the computer, and a fixture for placing a scanning object, the fixture having three small balls, the three small balls forming an inequilateral triangle, The computer includes: a point cloud acquisition module for acquiring a point cloud of an object obtained by scanning the scanner on each side of the object and a point cloud corresponding to three small balls on the jig of each side; a ball fitting module, A set of small balls are respectively fitted to the point clouds of the three small balls obtained from each side of the scanned object; a calculation module is used for calculating the position of each of the small balls in each group, and calculating each group The distance between each two small balls in the small ball; the matching module is used to take a certain surface of the object as a reference surface, and the matching small ball obtained by scanning the surface is used as a reference ball, and the other surface is found a matching ball for each small ball; and an alignment module for aligning each group of fitted balls obtained by scanning other faces of the object according to a matching relationship by a translational and rotational transformation to a fitting ball as a reference Position, get the transformation in the alignment process Array, and the point cloud of the object obtained by scanning the reference plane of the object is used as a reference point cloud, and the point cloud obtained by scanning each surface of the object is aligned to the position where the reference point cloud is located according to the obtained transformation matrix, and a complete point cloud of the object is obtained. . The discrete point cloud integration system of claim 1, wherein the computer further comprises a point cloud trimming module for trimming the matched point cloud of each small ball, and for obtaining After the complete point cloud of the object, delete the point cloud of the three small balls on the fixture obtained by fitting the small ball and scanning each side of the object. The discrete point cloud integration system of claim 1, wherein the computer further comprises a point cloud output module for outputting a complete point cloud of the aligned object and reporting the alignment accuracy. A discrete point cloud integration method, the method comprising the steps of: fixing an object to be scanned on a fixture that is reversible and containing three small balls, the three balls forming an inequilateral triangle; and flipping the fixture to scan the object All the faces of the object, while scanning each side of the object, scan the three balls on the jig once; obtain a point cloud of the object obtained by scanning each side of the object, and corresponding to the three small points on the jig a point cloud of the ball; a set of small balls are respectively fitted according to the point clouds of the three small balls obtained on each side of the scanned object; the positions of the small balls in each set of fitted small balls are calculated, and the fitting of each group is calculated. The distance between each two small balls in the ball; taking a certain surface of the object as a reference surface, using the fitted ball obtained by scanning the surface as a reference ball, finding a small match in the other faces matching the small balls on the face Ball; the set of fitted balls obtained by scanning other faces of the object are aligned according to the matching relationship by the translation and rotation transformation to the position of the fitting ball as the reference, and the transformation matrix in the alignment process is obtained; and the reference object is scanned. Obtained object Cloud point cloud as a reference, according to the obtained transformation matrix of each object surface obtained scanning point clouds are aligned to the reference cloud point, cloud point to obtain the complete object. The discrete point cloud integration method according to claim 4, further comprising the step of trimming the scrambled point clouds of each of the fitted small balls. The method for integrating a discrete point cloud according to claim 4, wherein the method further comprises the following steps: after obtaining a complete point cloud of the object, deleting the fixture obtained by fitting the small ball and scanning each surface of the object The point cloud of the ball. The discrete point cloud integration method according to claim 4, the method further comprising the steps of: outputting a complete point cloud of the aligned object and reporting the alignment accuracy.